Decoding the hallmarks of allograft dysfunction with a comprehensive pan-organ transcriptomic atlas.

Area of Science:

• Transplant immunology
• Genomics
• Bioinformatics

Background:

• Organ transplantation is vital, but allograft dysfunction remains a challenge.
• Existing 'omics' studies often focus on single organs, limiting a unified understanding of transplant pathology.
• Knowledge gaps hinder the distinction of pathological mechanisms across different transplanted organs.

Purpose of the Study:

• To comprehensively study human pan-organ allograft dysfunction using 'omics' data.
• To identify transcriptomic differences associated with allograft dysfunction, tolerance, and stable function across multiple organs.
• To develop and validate a cross-organ predictive model for allograft dysfunction.

Main Methods:

• Analysis of 150 datasets (>12,000 samples) from heart, lung, liver, and kidney transplants.
• Exploration of transcriptomic data to identify genes correlated with dysfunction (delayed graft function, acute rejection, fibrosis).
• Development of a transfer learning omics prediction framework for cross-organ analysis and validation in a kidney transplant cohort.

Main Results:

• Identification of genes robustly correlating with allograft dysfunction across all four studied organs.
• Demonstration that the transfer learning model achieved superior classification performance compared to single-organ models.
• Validation of the cross-organ predictive approach in a prospective kidney transplant cohort.

Conclusions:

• Machine learning models can effectively learn and generalize across different transplanted organs.
• The developed transcriptomic transplant resource can aid in developing pan-organ biomarkers for allograft dysfunction.
• This approach shows potential clinical utility for improving transplant outcomes through predictive modeling.